Abstract
The cause of colorectal cancer (CRC) is multifactorial, involving both genetic variants and environmental risk factors. We systematically searched the MEDLINE, EMBASE, China National Knowledge Infrastructure (CNKI) and Wanfang databases from inception to December 2016, to identify systematic reviews and meta‐analyses of observational studies that investigated gene–environment (G×E) interactions in CRC risk. Then, we critically evaluated the cumulative evidence for the G×E interactions using an extension of the Human Genome Epidemiology Network's Venice criteria. Overall, 15 articles reporting systematic reviews of observational studies on 89 G×E interactions, 20 articles reporting meta‐analyses of candidate gene‐ or single‐nucleotide polymorphism‐based studies on 521 G×E interactions, and 8 articles reporting 33 genome‐wide G×E interaction analyses were identified. On the basis of prior and observed scores, only the interaction between rs6983267 (8q24) and aspirin use was found to have a moderate overall credibility score as well as main genetic and environmental effects. Though 5 other interactions were also found to have moderate evidence, these interaction effects were tenuous due to the lack of main genetic effects and/or environmental effects. We did not find highly convincing evidence for any interactions, but several associations were found to have moderate strength of evidence. Our conclusions are based on application of the Venice criteria which were designed to provide a conservative assessment of G×E interactions and thus do not include an evaluation of biological plausibility of an observed joint effect.
Keywords: colorectal cancer, diet, environment, gene, interaction, risk factor
Abbreviations
- AICR
American Institute for Cancer Research
- CI
confidence interval
- CNKI
China National Knowledge Infrastructure
- CRC
colorectal cancer
- CTNNB1
cadherin‐associated protein β1
- CUP
continuous update project
- CYP24A1
cytochrome P450 family 24 subfamily A member 1
- G×E
gene–environment
- GATA3
GATA binding protein 3
- GECCO
Genetics and Epidemiology of Colorectal Cancer Consortium
- GWAS
genome‐wide association study
- HIATL1
hippocampus abundant transcript‐like 1
- MINK1
misshapen like kinase 1
- NAT2
N‐acetyltransferase 2
- NSAID
nonsteroidal anti‐inflammatory drug
- OR
odds ratio
- PI3K
phosphatidylinositol‐4,5‐bisphosphonate 3‐kinase
- PIK3C2G
phosphatidylinositol‐4‐phosphate 3‐kinase catalytic subunit type 2 gamma
- PTCHD3
patched domain containing 3
- RR
relative risk
- SHMT1
serine hydroxymethyltransferase 1
- SNP
single‐nucleotide polymorphism
- SOCCS
Scottish Colorectal Cancer Study
- TCF7L2
transcription factor 7 like 2
- WCRF
World Cancer Research Fund International
Introduction
Colorectal cancer (CRC) is the third most common cancer worldwide, with 746,000 new cases in men and 614,000 new cases in women.1 In some low‐to‐middle‐income countries, the incidence of CRC has been increasing partly due to changes in lifestyle and environment combined with aging populations.2 Thus, it is critical to understand both modifiable and non‐modifiable risk factors for CRC as this may enable more specific prevention strategies and risk assessment, especially in developing countries where CRC screening may not be feasible or affordable.3
CRC risk is determined by a complex interplay of both genetic variants and environmental exposures. It has been speculated that their interactions ‐ known as gene–environment (G×E) interactions should also be important determinants of CRC risk. To date, genome‐wide association study (GWAS) have shown that up to 50% of CRC heritability can be explained by common and rare variants included in popular genotyping arrays.4 Additional variants associated with CRC susceptibility that cannot be easily detected in GWAS by marginal effects of genetic factors may be identified by testing for interactions between single‐nucleotide polymorphisms (SNPs) and environmental risk factors.5, 6 Thus, identification of G×E interactions influencing CRC susceptibility may help to discover novel genetic and environmental risk factors for CRC, and extend our understanding of biological pathways and mechanisms of cancer etiology.
A number of systematic reviews, meta‐analyses and genome‐wide G×E interaction analyses that explored interaction effects in CRC have been published. We recently collected and evaluated the evidence across existing meta‐analyses of observational studies in dietary factors and gene‐diet interactions for the 5 most common cancers.7 Here, we performed an umbrella review to collect, update, and assess the evidence across existing systematic reviews, meta‐analyses and genome‐wide G×E interaction analyses that have explored the joint effects between genes and a wider range of environmental exposures in CRC. Our aim is to provide an overview on the associations between G×E interactions and CRC risk and to pinpoint which of the associations have robust evidence by evaluating the strength of the evidence using predetermined guidelines.
Methods
Search strategy
We systematically searched the MEDLINE, EMBASE, China National Knowledge Infrastructure (CNKI) and Wanfang databases from inception to December 2016. The search strategy and Medical Subject Headings terms are displayed in Supporting Information Table S1. All identified publications went through a 2‐step review before being included. Titles were reviewed by 1 investigator (TY). Abstracts and full texts were reviewed by 2 investigators (TY and MT). Any discrepancies were resolved by discussion.
Eligibility criteria
Three types of studies were eligible for this umbrella review: (i) systematic reviews of observational studies assessing interaction effects between genes and environmental exposures in CRC; (ii) meta‐analyses of candidate gene‐ or SNP‐based studies and analyses combining individual level data from multiple studies exploring G×E interactions and (iii) genome‐wide investigation of G×E interactions on CRC risk within GWAS consortia. We excluded reviews without explicit systematic literature searches; and systematic reviews or meta‐analyses of observational studies that explored associations between CRC risk and genes or environmental exposures only.
Data extraction
One investigator (TY) extracted data which were then checked by a second investigator (ZM). For each eligible article, we extracted the first author, year of publication, the dietary and genetic risk factors examined, study design and the number of studies included. For meta‐analyses, we extracted the summary study‐specific relative risk estimates [relative risk (RR), odds ratio (OR)] along with the corresponding 95% confidence intervals (CIs), the number of cases and total participants, the p value for interaction and the p value (or I 2) for heterogeneity.
Statistical analysis
For systematic reviews, we performed descriptive analyses and presented the authors’ main conclusions. The evaluation process is described in detail in Supporting Information methods.
For meta‐analyses of G×E interactions with a 2‐sided p value for interaction <0.05, or for G×E interactions reaching genome‐wide significance threshold, we used an extension of the Human Genome Epidemiology Network's Venice criteria8, 9 to evaluate the strength of the evidence (Table 1 10, 11, 12, 13, 14, 15, 16, 17). These guidelines have been used previously to assess cumulative evidence on joint effects of genes and environments on cancer risk.7, 18
Table 1.
Steps | Description |
---|---|
Step 1 Score for the strength of the observed evidence for the G×E interactions |
First, we scored the strength of the observed evidence for the interaction between environmental exposures and genetic variants. Each G×E association was graded based on (i) the amount of evidence, (ii) the extent of replication and (iii) the protection from bias. |
(i) For the amount of evidence, the grade A, B or C was assigned when the total number of individuals in the smallest comparison group (assuming 1:1 ratio of cases and controls) in the meta‐analysis was greater than 1,000, 100–1,000, or less than 100, respectively. | |
(ii) The replication consistency was assessed by the reported heterogeneity: grade A, I 2 < 25%; grade B, 25% ≤ I 2 ≤ 50%; grade C, I 2 > 50% or p value for heterogeneity <0.10. | |
(iii) For protection from bias 3 aspects of G×E association were taken into account as suggested by Boffetta P et al.8: protection from bias for the environmental exposure, for the genetic analysis and for the overall interaction. Grade A means that bias, if present, may change the magnitude but not the presence of an association; grade B means that there is no evidence of bias that would invalidate an association, but important information is missing; and grade C means that there is a strong possibility of bias that would render the finding of an association invalid. | |
On the basis of the combination of these 3 criteria (amount of evidence, degree of replication and protection from bias, each of which can be scored A, B and C), the epidemiological evidence for the association between G×E interaction and CRC risk was classified as strong, moderate or weak8 (Supporting Information Fig. S1). | |
Step 2 Prior score (expected) for G×E interactions |
Second, we established a prior score category (expected) for the G×E interactions using a framework presented in Boffetta P et al.8, which is based on prior scores for (i) the evidence of the main environmental and (ii) the evidence of the main genetic effects (Supporting Information Table 2). |
(i) Environmental main effect score: We scored the main environmental effects based on the meta‐analyses of the associations between environmental factors and CRC risk that were presented in the World Cancer Research Fund International (WCRF)/American Institute for Cancer Research (AICR) Third Expert Report,10 the subsequent Continuous Update Project (CUP) CRC reports11 and the CUP CRC Systematic Literature Review 2016.12 For the information of environmental risk factors that was not available in the above mentioned sections, we performed an additional literature search in MEDLINE and abstracted the relevant data as summarized and presented in Supporting Information methods. We then categorized the environmental factors in terms of strength of evidence by applying previously described set of criteria.13 The evidence was classified as convincing (Class I), highly suggestive (Class II), suggestive (Class III) or weak evidence (Class IV) based on sample size, highly significant summary associations, the 95% prediction intervals, presence of the small‐study effect and the excess significance bias. | |
(ii) Genetic main effect score: For the genetic main effects, a search in the National Human Genome Research Institute‐European Bioinformatics Institute catalog of GWAS,14 the GWAS central database15 and MEDLINE was conducted as described in Supporting Information methods. Subsequently, we scored the genetic associations using the Human Genome Epidemiology Network Venice criteria.9, 16, 17 Only genetic effects with p <10−5 were considered for evaluation, and the evidence was classified as strong, moderate or weak based on a combination of the 3 criteria (amount of evidence, degree of replication and protection from bias), each of which was scored A, B or C (Supporting Information Fig. S1). For the genetic variants that reached genome‐wide significance threshold, the evidence class of the genetic variant was only based on the amount of evidence.16 The search strategies, the Medical Subject Headings terms and the numbers of hits are presented in Supporting Information Tables 3 and 4, respectively. | |
Step 3 Combined score |
Lastly, we examined the overall plausibility of each interaction by combining the prior score and the strength of the observed evidence. Higher weight was given to the observed evidence in case of conflicting results between the observed evidence and the prior scores. |
First, we scored the strength of the observed evidence for the interaction between environmental exposures and genetic variants (observed score). Each G×E association was graded based on the amount of evidence, the extent of replication and the protection from bias (Table 1). On the basis of the combination of these 3 criteria, the epidemiological evidence for the association between G×E interaction and CRC risk was classified as strong, moderate or weak8 (Supporting Information Fig. S1).
Second, we established a prior score category (expected) for the G×E interactions using a framework presented in Boffetta et al.,8 which is based on prior scores for (i) the evidence of the main environmental and (ii) the evidence of the main genetic effects (Table 1). In brief, we established the prior score for the interactions based on the strength of evidence for the main environmental effect and the main genetic effect (1 = strong, 2 = moderate, 3 = weak) (Supporting Information Table S2). When both of the evidence were convincing (Class I), then the prior score category was strong. When one of the evidence was suggestive (Class III) or weak (Class IV), then the prior score category was weak.
Third, we examined the overall plausibility of each interaction by combining the prior score and the strength of the observed evidence. Higher weight was given to the observed evidence in case of conflicting results between the observed evidence and the prior scores.
Finally, for the statistically significant G×E interactions (with a 2‐sided p value for interaction <0.05) that were identified from the candidate meta‐analyses or for interactions that were concluded as suggestive by the authors of the systematic reviews, we tested the interactions in the Scottish Colorectal Cancer Study (SOCCS)19 dataset, and we also compared to results from the Genetics and Epidemiology of Colorectal Cancer Consortium (GECCO).20
Results
Number and type of articles identified
Overall, 14,219 publications were identified across the 4 databases. After applying the inclusion and exclusion criteria, 42 publications were selected for inclusion (1 was in Chinese; Fig. 1). The details of 89 G×E interactions covering 22 environmental exposures identified in 15 systematic reviews of observational studies21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35 are presented in Supporting Information Results section.
Main findings of meta‐analyses of candidate gene‐ or SNP‐based studies
Twenty articles21, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54 reporting meta‐analyses of candidate gene‐ or SNP‐based studies and analyses combining individual level data from multiple studies explored G×E interactions on CRC risk, covering 20 environmental exposures and 43 genes or genetic variants (Supporting Information Table S5). We identified 5 G×E interactions with 2‐sided p value for interaction <0.05 (or adjusted p <0.05 after accounting for multiple comparisons): N‐acetyltransferase 2 (NAT2) and processed meat intake51; NAT2 and red meat intake51; rs16892766 (8q23.3) and vegetable consumption39; serine hydroxymethyltransferase 1 (SHMT1) C1420T polymorphism and folate intake46; and rs6983267 (8q24) and aspirin use44 (Supporting Information Table S5). Also, the interactions between the above 5 environmental exposures (processed meat, red meat, vegetables, folate, aspirin use) and approximately 2.7 million genetic variants for CRC risk were also explored in GWAS consortia55, 56, 57 (Supporting Information Tables S6 and S7). However, none of the interactions observed in the candidate‐based studies were detected at the genome‐wide significance level in the GWAS consortia. We also tested interactions between rs16892766 and vegetable consumption, SHMT1 C1420T polymorphism and folate intake and rs6983267 and aspirin use in the SOCCS,19 and compared to the results from the GECCO.20 None of the associations were nominally significant (α = 0.05) in our data (data not presented) or in the GECCO (data not presented). Each identified G×E interaction is described in detail in Supporting Information results.
Main findings of genome‐wide investigation of G×E interactions within GWAS consortia
Eight articles55, 56, 57, 58, 59, 60, 61, 62 corresponding to 33 genome‐wide G×E interaction analyses explored joint effects between a large number of common polymorphisms and 22 selected environmental exposures on CRC risk by using agnostic searches (a summary of all the analyses and details are presented in Supporting Information Tables S6; SNPs with the smallest p value for interaction from each genome‐wide G×E interaction analysis are presented in Supporting Information Table S7). The following G×E interactions were identified that reached genome‐wide significance threshold specified by the authors of the original publication: rs4143094 (10p14) and processed meat intake55; rs9409565 (9q22.32) and light‐to‐moderate alcohol drinking (1–28 g/day)60; rs2965667 (12p12.3), rs16973225 (15q25.2) and aspirin and/or nonsteroidal anti‐inflammatory drug (NSAID) use57; patched domain containing 3 (PTCHD3) at 10p12.1, misshapen like kinase 1 (MINK1) at 17p13.2 and NSAID use61; rs964293 (20q13.2) and use of estrogen plus progestogen therapy59; and rs1944511 (11q23.3) and overweight62 (Supporting Information Table S7). Each identified genome‐wide G×E interaction is described in detail in Supporting Information results.
Evaluation of the evidence for G×E interactions with main genetic effects of a p < 10−5
Here, we present the evidence for the identified G×E interactions in relation to CRC risk with main genetic effects (p < 10−5; Fig. 2).
Gene‐aspirin use interactions
Aspirin use was associated with a reduced risk of CRC on the basis of a meta‐analysis of 39 studies with 151,367 cases [users versus non‐users RR,0.79 (95% CI: 0.74, 0.85); p = 7.8×10−11; I 2 = 91.1%],63 thus the association was graded as highly suggestive (class II) due to the high heterogeneity between the studies (Table 2). The main effect of rs6983267 (8q24) on CRC risk was graded as strong (ABA, equivalent to AAA based on the Venice criteria16) in a meta‐analysis including 13,348 cases and 26,438 controls of European ancestry [OR, 0.84 (95% CI: 0.80, 0.88); p = 7.45×10−13; I 2 = 37.7%]64 (Table 3). Consequently, the interaction between rs6983267 (8q24) and aspirin use was given a moderate prior score (Moderate – 2) and a moderate overall plausibility score (Table 4).
Table 2.
Environmental exposure | Reference | Unit of comparison | Number of cases | Number of studies | Relative risk (95% CI) | p Value | Prediction interval | Heterogeneity (I 2 and/or p value) | Evidence class1 |
---|---|---|---|---|---|---|---|---|---|
Processed meat | WCRF, CUP 2017 | Per 50 g/day | 10,738 | 10 | 1.16 (1.08, 1.26) | 0.0002 | NA |
I
2
= 20.1%;
p = 0.258 |
III |
Red meat | WCRF, CUP 2017 | Per 100 g/day | 6,662 | 8 | 1.12 (1.00, 1.25) | 0.05 | NA |
I
2
= 23.6%;
p = 0.241 |
IV |
Light‐to‐moderate alcohol intake | WCRF, CUP 2017 | 1 drink/day | 36,942 | 8 | 1.06 (1.00, 1.11) | 0.03 | NA |
I
2
= 60.4%;
p = 0.013 |
IV |
Vegetables | WCRF, CUP 2017 | 100 g/day | 14,136 | 11 | 0.98 (0.96, 0.99) | 0.01 | NA |
I
2
= 0.0%;
p = 0.48 |
IV |
Total folate | WCRF, CUP 2017 | 100 mcg/day | 4,633 | 8 | 0.99 (0.98, 1.00) | 0.05 | NA |
I
2
= 0.0%;
p = 0.92 |
IV |
Aspirin use | Qiao Y, 2018 | Users vs. non‐users | 151,367 | 39 | 0.79 (0.74, 0.85) | 7.8×10−11 | NA |
I
2
= 91.1%; p = 0.000 |
II2 |
Aspirin and/or NSAID use | Nan H, 2015 | Regular users of aspirin, NSAIDs or both vs. non‐regular users | 8,634 | 10 | 0.69 (0.64, 0.74) | 6.20×10−28 | NA | p = 0.02 | II2 |
Use of estrogen plus progestogen therapy | Lin KJ, 2012 | Ever users vs. non‐users | NR | 17 | 0.74 (0.68, 0.81) | <0.001 | NA |
I
2
= 0%;
p = 0.88 |
III |
Overweight | WCRF, CUP 2017 | Per 5 kg/m2 increase in BMI | 71,089 (total number) | 38 (20 for men and 24 for women) | Men: 1.08 (1.04, 1.11); women: 1.05 (1.02, 1.08) |
Men: p < 0.001 Women: p < 0.001 |
NA |
Men: I
2
= 83.3%, p < 0.001; women: I 2 = 82.5%, p < 0.001 |
Men: III; Women: III |
Abbreviations: BMI, body mass index; CI, confidence interval; CUP, Continuous Update Project; G×E, gene–environment; NA, not applicable; NR, not reported; NSAID, nonsteroidal anti‐inflammatory drug; vs., versus; WCRF, World Cancer Research Fund.
Evidence class was decided using the after criteria: Convincing evidence (class I) required >1,000 cases, highly significant summary associations (p < 10−6 by random effects), a 95% prediction interval not including the null, no evidence of small‐study effects, no evidence of excess significance bias, and low heterogeneity values (I 2 < 50%). Highly suggestive evidence (class II) required >1,000 cases, highly significant summary associations (p < 10−6 by random effects), and the largest study to have a 95% confidence interval that excluded 1. Suggestive evidence (class III) required only >1,000 cases and p < 0.001 by random effects. Evidence was considered weak (class IV) when p < 0.05. No association indicates evidence for the main environmental effects with p > 0.05.
The evidence was classified as highly suggestive (class II) due to the high heterogeneity between the studies.
Table 3.
Genetic variant | Gene (or near gene) | Reference | Discovery sample size | Replication sample size / Number of studies in meta‐analysis | Reported OR (95% CI) | p Value for genetic main effect | Heterogeneity, I 2 | Venice criteria | Evidence class1 |
---|---|---|---|---|---|---|---|---|---|
rs4143094 | 10p14/GATA3 | Figueiredo JC, 2014 | 9,287 cases and 9,117 controls of European ancestry from USA, Australia, Canada and Germany | Meta‐analysis, 10 studies | NR | 0.26 | NR | NA | No association |
Slow/intermediate/rapid | NAT2 | Zhang L, 2012 | 13,606 cases and 17,957 controls of Africans, Asians, Caucasians and mixed populations | Meta‐analysis, 39 studies | Slow vs. rapid phenotype: 0.96 (0.90, 1.01) | No association | I 2 = 17.8% | NA | No association |
Slow/intermediate/rapid | NAT2 | Wang H, 2015 | 2,186 cases and 3,736 controls of Japanese; 466 cases and 4,356 controls of African Americans | Meta‐analysis, 7 studies | Rapid vs. slow phenotype: Japanese: 1.05 (0.87, 1.27); African Americans: 0.75 (0.50, 1.14); Combined: 0.99 (0.83, 1.18) |
Japanese: 0.77; African Americans: 0.19; Combined: 0.81 |
NR | NA | No association |
rs9409565 | 9q22.32/HIATL1 | Schumacher FR, 20152 | 18,299 cases and 19,656 controls of European ancestry from North America, Australia and Europe | Meta‐analysis, 4,725 cases and 9,969 controls of East Asian ancestry from Republic of Korea, China and Japan | 0.98 (0.95, 1.01) | 0.127 | NR | NA | No association |
rs16892766 | 8q23.3/EIF3H | Li M, 2015 | 41,728 cases and 44,393 controls | Meta‐analysis, 11 studies | 1.22 (1.18, 1.27) | 1.39 × 10−24 | I 2 = 4% | AAA | Strong |
rs6983267 | 8q24.21 | Tanskanen T, 2017 | 1,701 Finnish cases and 14,082 population‐based, cancer‐free controls | Meta‐analysis, 13,348 cases and 26,438 controls of European ancestry | 0.84 (0.80, 0.88) | 7.45 × 10−13 | I 2 = 37.7% | ABA (equivalent to AAA) | Strong |
C1420T | SHMT1 | Wang Q, 2014 | 3,912 cases and 4,954 controls | Meta‐analysis, 7 studies | TT vs. CC: 0.84 (0.73, 0.97); CT vs. CC: 1.01 (0.92, 1.10); TT + CT vs. CC: 0.97 (0.89, 1.06); TT vs. CT + CC: 0.84 (0.73, 0.96) |
TT vs. CC: 0.020; CT vs. CC: 0.903; TT + CT vs. CC: 0.476; TT vs. CT + CC: 0.013 |
TT vs. CC: I
2 = 3.8%; CT vs. CC: I 2 = 0%; TT + CT vs. CC: I 2 = 0%; TT vs. CT + CC: I 2 = 0% |
NA | No association |
rs2965667 | 12p12.3/PIK3C2G | Orlando G, 2016 | 8,749 cases and 18,245 controls from Europe | Meta‐analysis, 7 studies | 0.97 (0.87, 1.08) | 0.552 | I 2 = 4.8% | NA | No association |
rs16973225 | 15q25.2/interleukin 16 | Orlando G, 2016 | 8,749 cases and 18,245 controls from Europe | Meta‐analysis, 7 studies | 1.05 (0.97, 1.15) | 0.242 | I 2 = 0% | NA | No association |
rs964293 | 20q13.2/CYP24A1 | Orlando G, 2016 | 8,749 cases and 18,245 controls from Europe | Meta‐analysis, 7 studies | 0.97 (0.93, 1.01) | 0.156 | I 2 = 6.3% | NA | No association |
Including 7 variants3 | 10p12.1/PTCHD3 | Timofeeva M, 2015 | 8,100 cases and 21,820 controls from Europe | Meta‐analysis, 6 studies | NR | 0.352 | NR | NA | No association |
Including 8 variants4 | 17p13.2/MINK1 | Timofeeva M, 2015 | 8,100 cases and 21,820 controls from Europe | Meta‐analysis, 6 studies | NR | 0.381 | NR | NA | No association |
rs1944511 | 11q23.3 | Siegert S, 2013 | 259 cases and 1,002 controls | Genome‐wide G×E interaction analysis | 1.07 | 0.536 | NR | NA | No association |
Abbreviations: CI, confidence interval; G×E, gene–environment; NA, not applicable; NR, not reported; OR, odds ratio; vs., versus.
Evidence class was decided on the basis of the Human Genome Epidemiology Network's Venice criteria: No association indicates evidence for main genetic effects with p > 10−5. Only genetic effects with p < 10−5 were considered for evaluation. On the basis of a combination of 3 criteria (amount of evidence, degree of replication, and protection from bias) (each of which can be scored A, B, or C), the epidemiological evidence for an effect of the genotype is classified as strong, moderate, or weak. For amount of evidence, a grade of A, B, or C was assigned when the sample size for the rarer genotype in the meta‐analyses was greater than 1,000, 100–1,000, or less than 100, respectively. For replication consistency, we used I 2 < 25% to assign grade A, 25–50% to assign grade B, and > 50% or a p value for heterogeneity <0.10 to assign grade C. For protection from bias, a grade of A means that bias, if present, may change the magnitude but not the presence of an association; a grade of B means that there is no evidence of bias that would invalidate an association, but important information is missing; and a grade of C means that there is a strong possibility of bias that would render the finding of an association invalid. For the genetic variants that reached genome‐wide significance threshold, the evidence class of the genetic variant was only based on the amount of evidence based on the clarification of Venice Criteria (Khoury MJ et al, 2009).
Current study exploring the marginal association of rs9409565 was used since it is about twice as large as Gong et al. (2016).
Seven variants at 10p12.1 were included in the analysis that explored main genetic effects by Timofeeva M et al, 2015. However, the interaction analysis (by Jiao S et al, 2015) included 8 variants: chr10:27687284, chr10:27687437, chr10:27687638, chr10:27687775, chr10:27687989, chr10:27688101, chr10:27702174 and chr10:27702624.
Eight variants at 17p13.2 were included in the analysis that explored main genetic effects Timofeeva M et al, 2015. However, the interaction analysis (by Jiao S et al, 2015) included 4 variants: chr17:4794313, chr17:4794407, chr17:4796839 and chr17:4797910.
Table 4.
Environmental exposure | Genetic variant | Gene (or near gene) | Score based on observed evidence1 | Strength of observed evidence for interaction | Score for environmental evidence (evidence class) | p Value for main genetic effect | Score for genetic evidence /Venice criteria2 | Prior score3 | Combined score4 |
---|---|---|---|---|---|---|---|---|---|
Meta‐analyses of candidate gene‐ or SNP‐based studies | |||||||||
Processed meat | Slow/intermediate/rapid | NAT2 | (B or C) BC | Weak | III | No association | No association | Weak: 3 | No evidence |
Red meat | Slow/intermediate/rapid | NAT2 | (B or C) AC | Weak | IV | No association | No association | Weak: 3 | No evidence |
Vegetables | rs16892766 | 8q23.3/EIF3H | CBB | Weak | IV | 1.39 × 10−24 | Strong/AAA | Weak: 3 | No evidence |
Folate intake | C1420T | SHMT1 | CAC | Weak | IV | TT vs. CC: 0.020; CT vs. CC: 0.903; TT + CT vs. CC: 0.476; TT vs. CT + CC: 0.013 | No association | Weak: 3 | No evidence |
Aspirin use | rs6983267 | 8q24 | BAB5 | Moderate | II | 7.45 × 10−13 | Strong/ABA (equivalent to AAA) | Moderate: 2 | Moderate |
Genome‐wide G×E interaction analyses | |||||||||
Processed meat | rs4143094 | 10p14/GATA3 | BBB | Moderate | III | 0.26 | No association | Weak: 3 | Moderate |
Light‐to‐moderate drinking | rs9409565 | 9q22.32/HIATL1 | BBA | Moderate | IV | 0.127 | No association | Weak: 3 | Moderate |
Aspirin and/or NSAID use | rs2965667 | 12p12.3/PIK3C2G | B ‐ A | Moderate | II | 0.552 | No association | Weak: 3 | Moderate |
rs16973225 | 15q25.2/interleukin 16 | B ‐ A | Moderate | II | 0.242 | No association | Weak: 3 | Moderate | |
NSAID use | Including 8 variants | 10p12.1/PTCHD3 | ‐ ‐ B | Not possible to evaluate6 | II | 0.352 | No association | Weak: 3 | Not possible to evaluate |
Including 4 variants | 17p13.2/MINK1 | ‐ ‐ B | Not possible to evaluate6 | II | 0.381 | No association | Weak: 3 | Not possible to evaluate | |
Use of estrogen plus progestogen therapy | rs964293 | 20q13.2/CYP24A1 | BBA | Moderate | III | 0.156 | No association | Weak: 3 | Moderate |
Overweight | rs1944511 | 11q23.3 | C ‐ B | Weak | III | 0.536 | No association | Weak: 3 | No evidence |
Abbreviation: CRC, colorectal cancer; G×E, gene–environment; NSAID, nonsteroidal anti‐inflammatory drug; SNP, single‐nucleotide polymorphisms; vs., versus.
The strength of the observed evidence for interaction between the environmental exposures and the genetic variants was based on an extension of the Human Genome Epidemiology Network's Venice criteria used for assessing cumulative evidence for genetic associations. Each G×E association was graded based on the amount of evidence, the extent of replication and protection of bias. Dashes indicate that 1, 2, or 3 elements of the Venice criteria cannot be decided. A complete score should have 3 letters, corresponding to amount of evidence, degree of replication, and protection from bias components of the Venice criteria. If 1 element is missing, the score is represented by a single dash and 2 letters. If 2 elements are missing, the score is represented by 2 dashes and a letter.
No association indicates evidence for main genetic effects with p > 10−5. Only genetic effects with p < 10−5 were considered for evaluation using the Human Genome Epidemiology Network's Venice criteria.
The prior score was based on scores for environmental evidence and genetic evidence (Table 2 and Table 3).
The overall plausibility of an interaction was examined by comparing the prior score and the score for the strength of the observed evidence. Higher weight was given to the observed evidence in case of conflicting results between the prior and observed scores.
The replication consistency was graded as A because the interaction between aspirin use and rs6983267 was replicated in the GECCO.
Jiao et al (2015) presented no information on the total number of individuals in the smallest comparison group and heterogeneity between the studies. Therefore evaluating the amount of evidence and the extent of replication according to the Venice criteria was not possible.
Gene‐vegetable interactions
A meta‐analysis of the association between vegetable intake and CRC risk in the latest Continuous Update Project (CUP) CRC reports11 and the CUP Colorectal Systematic Literature Review 201612 showed a reduction in CRC risk with 100 g/day increase in vegetable intake [RR, 0.98 (95% CI: 0.96, 0.99); p = 0.01; I 2 = 0%; n = 11 prospective studies; n of cases = 14,136] and the association was classified as class IV (weak) (Table 2). On the basis of the strong (AAA, based on the Venice criteria16) main genetic (Table 3) and weak (class IV) environmental effects of vegetable intake on CRC risk, the possible interactions between 8q23.3 locus and vegetable intake on CRC risk was given a weak (Weak – 3) prior score and therefore no evidence was found for this interaction7 (Table 4).
Evaluation of the evidence for G×E interactions with no main genetic effects (p > 10−5)
Here, we present the evidence for the identified G×E interactions in relation to CRC risk with no main genetic effects (p > 10−5; Fig. 2). The G×E interactions were considered as tenuous even if they were classified as having moderate evidence.
Interactions between genetic variants and use of aspirin, NSAIDs or both
In a meta‐analysis of 10 studies including 8,634 cases and 8,553 controls, regular use of aspirin and/or NSAIDs, compared to non‐regular use, was associated with lower risk of CRC [RR 0.69 (95% CI: 0.64, 0.74); p = 6.20 × 10−28; p heterogeneity = 0.02], thus the evidence was classified as highly suggestive (class II) due to the high heterogeneity between the studies57 (Table 2). For the main genetic effects, no associations were observed between rs2965667 (12p12.3), rs16973225 (15q25.2) and CRC risk (p > 10−5) in a meta‐analysis of 7 GWAS from Europe including 8,749 cases and 18,245 controls (p = 0.552 and 0.242, respectively)65 (Table 3). Thus, the interactions between rs2965667 (12p12.3), rs16973225 (15q25.2) and aspirin and/or NSAID use were given moderate overall plausibility scores and weak (Weak – 3) prior scores (Table 4). Nevertheless, the overall plausibility scores for the interactions between 10p12.1/PTCHD3, 17p13.2/MINK1 and NSAID use could not be properly evaluated due to the missing elements of the extension of the Venice criteria8 that was used for assessing the observed evidence for the interactions (Table 4).
Interactions between genetic variants and use of estrogen plus progestogen therapy
The RR for use of estrogen plus progestogen therapy on CRC risk was 0.74 (95% CI: 0.68, 0.81; p < 0.001; I 2 = 0%) in a meta‐analysis of 17 studies,66 thus the association was classified as suggestive (class III; Table 2). Furthermore, no association was observed between the rs964293 variant and CRC risk (p > 10−5) in the meta‐analysis of 7 GWAS [OR, 0.97 (95% CI: 0.93, 1.01); p = 0.156; I 2 = 6.3%]65 (Table 3). On the basis of the prior and observed scores, the interaction between rs964293 (20q13.2) and use of estrogen plus progestogen therapy was given a moderate overall plausibility score and a weak (Weak – 3) prior score (Table 4).
Gene‐alcohol interactions
A meta‐analysis of 8 prospective studies in the latest CUP CRC reports11 and the CUP Colorectal Systematic Literature Review 201612 showed one drink per day increase was associated with increased CRC risk [RR, 1.06 (95% CI: 1.00, 1.11); p = 0.03; I 2 = 60.4%; n of cases = 36,942], and the association between light‐to‐moderate drinking and CRC risk was categorized as weak (class IV) (Table 2). Additionally, no main effect was observed for the rs9409565 (9q22.32) on CRC risk (p > 10−5) of 18,299 cases [OR, 0.98 (95% CI: 0.95, 1.01); p = 0.127]67 (Table 3). Hence, the interaction between rs9409565 (9q22.32) and light‐to‐moderate drinking was given a weak prior score (Weak – 3) and a moderate overall plausibility score (Table 4).
Gene‐meat interactions
No evidence was found for the interactions between processed meat, red meat and NAT2 based on the weak (Weak – 3) prior score and the weak observed score (Tables 2, 3, 4). The possible 10p14 locus‐processed meat interaction was given a weak prior score (Weak – 3) and a moderate plausibility score7 (Tables 2, 3, 4).
Gene‐folate interactions
No evidence was found for the interaction between SHMT1 C1420T and folate intake on CRC risk based on a weak (Weak – 3) prior score and a weak observed score (Tables 2, 3, 4).
Gene‐overweight interactions
No evidence was found for the interaction between rs1944511 (11q23.3) and overweight based on a weak (Weak – 3) prior score and a weak observed score (Tables 2, 3, 4).
Discussion
Main findings
Based on the prior and observed scores, only the interaction between rs6983267 (8q24) and aspirin use was found with a moderate overall plausibility score and a main genetic effect (p = 7.45 × 10−13; strong, based on the Venice criteria). In particular, the benefit of regular aspirin use on CRC risk was confined to individuals with T allele of rs6983267, which has been associated with impaired binding of cadherin‐associated protein β1 (CTNNB1)/ transcription factor 7 like 2 (TCF7L2) and lower expression of MYC.68, 69, 70 Moreover, aspirin has been associated with Wnt pathway and the inhibition of nuclear CTNNB1 expression in colon cancer cell lines.71 Hence, it is suggested that a genetic background by which CTNNB1/TCF7L2 binding is not constitutively active is necessary for the susceptibility to the effects of aspirin on the Wnt/CTNNB1 pathway.44
Moderate overall plausibility scores were also found for the interactions between rs4143094 (10p14) and processed meat intake, rs9409565 (9q22.32) and light‐to‐moderate alcohol drinking (1–28 g/day), rs964293 (20q13.2) and use of estrogen plus progestogen therapy, as well as rs2965667 (12p12.3), rs16973225 (15q25.2) and aspirin and/or NSAID use. However, these interactions are regarded as tenuous due to the lack of main genetic effects (p > 10−5), even though they may provide clues to discovering novel CRC susceptibility loci that have not been readily detected in GWAS by their marginal effects of genetic factors.5, 6
Little is known about the underlying molecular mechanisms of the interactions between rs2965667 at the 12p12.3/ phosphatidylinositol‐4‐phosphate 3‐kinase catalytic subunit type 2 gamma (PIK3C2G) locus, rs16973225 (15q25.2/interleukin 16) and aspirin and/or NSAID use on CRC risk. PIK3C2G gene encodes a protein of the phosphatidylinositol‐4,5‐bisphosphonate 3‐kinase (PI3K) family,72 of which the activated signaling can inhibit apoptosis in colon cancer cell lines that can be restored with NSAID‐mediated blockade of PI3K.57, 73 Interleukin 16 may stimulate monocyte induction of proinflammatory cytokines associated with tumorigenesis,74, 75 which suggests that polymorphisms in or near the interleukin 16 gene may be associated with the production of inflammatory cytokines that modify the chemopreventive effect of aspirin and/or NSAIDs on CRC.57 However, it is proposed that those GWAS‐identified promising loci outside of a known coding region may affect more distant genes rather than the closest gene.76
The interaction between use of estrogen plus progestogen hormone preparations and the rs964293 variant in the 20q13.2, known as cytochrome P450 family 24 sub‐family A member 1 (CYP24A1) is biologically plausible. CYP24A1 is greatly expressed in malignant colon tumours,77 and some variants in CYP24A1 have been associated with CRC risk.78 In a recent meta‐analysis, ever use of estrogen plus progestogen therapy has been associated with lower CRC risk [OR: 0.74 (95% CI: 0.68, 0.81); p <0.001].66 Also, an inverse association was found in the meta‐analysis of randomized controlled trials with a hazard ratio of 0.77 (95% CI: 0.59, 0.98) (p = 0.037).66 It is suggested that CYP24A1 may only be a metabolizing enzyme for progestrogens but not estrogen, since an interaction effect was only found with the use of combined estrogen‐progestogen therapy and not with estrogen‐only intake.59
The mechanism of the modifying effect of the rs4143094 variant at the 10p14 locus, near GATA binding protein 3 (GATA3) region, on the association between processed meat intake and CRC risk is even less clear. GATA3 has been described as a master regulator of T‐helper 2 cell differentiation in mature CD4 (+) T cells and has been associated with T cell development.79 One possible explanation of the functional impact is that processed meat could trigger a pro‐tumorigenic inflammatory or immunological response,55, 80 and loss of GATA genes or silencing of expression can increase CRC risk.81
Furthermore, the mechanism of the modifying effect of alcohol consumption on the association between rs9409565 at the 9q22.32/ Hippocampus Abundant Transcript‐Like 1 (HIATL1) locus and CRC risk has not been understood. HIATL1 is a member of the solute carrier group of membrane transport, which makes the move of substances (such as peptides, amino acids, proteins, metals, and neurotransmitters) directly into or out of cells possible.82, 83 Gene expression analyses indicate that the variants at 9q22.32/HIATL1 that interact with alcohol on CRC risk through genome‐wide G×E interaction analyses can also impact HIATL1 expression,60 which suggests that alcohol may modify the effects of HIATL1 on CRC risk through its influence on HIATL1 expression levels.60
Challenges for G×E interaction studies
Studies of G×E interactions require much larger sample sizes than main effect analyses due to small effect sizes, multiple testing, misclassification due to imperfect measures of environmental exposures and more model parameters.5, 84, 85, 86 Even if the sample size is large enough to detect interactions with common exposures, it may still be insufficient when analyzing relatively rare exposures or genotypes of interest.5 Meta‐analyses of existing G×E interactions can address this sample size limitation, though investigators should be aware of other issues, such as i) inconsistencies in the definitions of exposures and outcomes and ii) differences in study designs, tools for assessing exposures, distributions of exposures, statistical analyses, presentation of results and publication bias.87 Additionally, the scale and distribution of environmental exposures in a population can also influence power. In dietary studies, it is suggested that the diet under investigation should be sufficiently variable in the population to allow evaluation of various intakes, and risk should be reported per similar units of exposure (e.g. per 100 g meat intake per day) to ensure that comparisons between populations is possible.21 In studies performed under the framework of GWAS consortia, the nature of exposure assessment may be different from other studies. Thus, harmonization of exposure may have been more difficult. Consequently, an increased sample size and decreased quality of exposure data, as well as a fully validated design may help to address the measurement error issue.5
Strengths and limitations of current review
The strengths of umbrella reviews have been described elsewhere.13, 88, 89 In our study, we found moderate evidence for some G×E interactions on CRC risk, though most of these interaction effects were tenuous due to the lack of main genetic effects and/or environmental effects. The proposed biological mechanisms for the G×E interactions are hypothetical and in the absence of experimental studies could not be used to prove causality. Ideally, evidence from a model system or/and organism with genetic variations in the gene/polymorphisms of interest and exposed to the physiological dose of the environmental factor (aspirin, processed meat, alcohol, and sex hormone) is required to support the epidemiological evidence described here. Thus, further replication and functional studies are required to confirm our findings and understand the biologic implications of the interactions.
Our study has limitations. First, interaction effects that have not yet been assessed through meta‐analyses or systematic reviews would not have been included, since umbrella reviews do not focus on individual studies. Second, we did not use an established tool to assess the risk of bias in the included observational studies, because available tools such as Q‐Genie90 and the Newcastle‐Ottawa Scale do not capture aspects relevant to the G×E assessment. Third, we excluded reviews without explicit systematic literature searches in order to avoid bias, but this could have resulted in the exclusion of syntheses of literature that have not been systematic.91 Fourth, interactions with limited evidence or limited sample size may have led to false‐negative findings for some joint effects that have long been thought to exist. False‐positives may also exist, although we have applied our criteria to assess the evidence to minimize biases. In our study, we used I 2 to assess the heterogeneity in the evaluation, however, it has been reported that I 2 represents what proportion of the observed variance would remain if we could eliminate the sampling error rather than how much the effect size varies.92 Additionally, we would miss interactions in which there were no marginal effects of genotype or exposure on CRC risk, since the Venice criteria only aim to grade the credibility of evidence of significant main effects; and 2 of the identified G×E interactions in our study could not be properly evaluated due to lack of information required to apply these criteria.
We combined systematic literature reviews, candidate and genome‐wide G×E studies together. The main reason for this was to provide a comprehensive overview of the existing literature on G×E interaction in relation to CRC risk. However, an issue is that in each of these types of studies, different criteria and p value thresholds are typically used to evaluate the creditability of findings. Candidate gene studies with liberal p value thresholds might be biased toward false positive findings, while the genome‐wide approach is prone to false negative observations, because true interactions may not reach a stringent genome‐wide significance threshold. This, as well as other issues such as genotyping and imputation problems, could be a reason why none of the interactions identified in the candidate gene studies were replicated in genome‐wide G×E studies.
The extension of Venice criteria used here has been applied in the past to assess joint effects of environments and genes on risk of multiple cancers.7, 10 This method does not take into account other lines of evidence, such as biological plausibility of observed associations, biological gradient of effects, coherence of the observations across multiple type of studies or support by experiments. Although other guidelines are more comprehensive in the range of evidence considered such as the newly developed “integrative research” method that combines causal criteria of Austin Bradford Hill with graphical models,93 we did not use it to evaluate the evidence in our study, since some of the criteria are difficult to apply and interpret in molecular epidemiology (e.g. temporality).94
Conclusions
Our assessment maps the status of evidence on the associations between G×E interactions and CRC risk. Despite the identified studies exploring a wide variety of G×E interactions on CRC, we conclude that we did not find highly convincing evidence for any interactions, but several associations were found to have moderate strength of evidence using our set of guidelines.
Though most of the evaluated G×E interactions in our study were with no main genetic effects, it has been suggested that such kind of risk loci that have not been readily detected in traditional GWAS may be identified by testing for interactions between SNPs and environmental risk factors, even though there is no strong evidence for a G×E interaction.5, 6 Thus, studies incorporating accurate assessment of environmental exposures are encouraged not only to identify novel G×E interactions, but also to discover novel risk loci for CRC by characterizing any underlying G×E interactions.5 Moreover, there remains insufficient evidence for G×E interactions on CRC risk, and some G×E interactions without strong evidence may still be important in CRC prevention. Hence, studies with large sample sizes and further functional studies are required to identify important G×E interactions that could have public health impact, so as to shed light to CRC etiology and to allow for more specific risk assessment for early‐detection or prevention strategies.
Conflict of interest
None declared.
Ethical approval
Not required.
Data sharing
No additional data available.
Authors’ contributions
Study design: MT and ET; Literature search: TY; Study selection: TY and MT; Data extraction: TY and ZM; Data analysis: TY and MT; Study draft and revision: TY, MT, ET, XL, ZM, JL, SMF, JPAI, MGD and HC; Article guarantor: Dr. Maria Timofeeva and Dr. Evropi Theodoratou.
Supporting information
Acknowledgements
The work was supported by funding for the infrastructure and staffing of the Edinburgh CRUK Cancer Research Centre. Acknowledgments for the Scottish Colorectal Cancer Study: We thank the participants in all of the studies that contributed to this piece of work and all the recruitment teams and collaborators who make such studies possible. We acknowledge the excellent technical support from Marion Walker. We are grateful to Ruth Wilson, Donna Markie and all those who continue to contribute to recruitment, data collection and data curation for the Study of Colorectal Cancer in Scotland studies. In addition to all consultant colorectal surgeons who provided data on their patients, we are also indebted to the chairs and offices of the managed clinical networks throughout Scotland who contributed substantially to clinicopathologic data and staging information. We acknowledge the expert support on sample preparation from the Genetics Core of the Edinburgh Wellcome Trust Clinical Research Facility.
Acknowledgements for the GECCO consortium: ASTERISK: We are very grateful to Dr. Bruno Buecher without whom this project would not have existed. We also thank all those who agreed to participate in our study, including the patients and the healthy control persons, as well as all the physicians, technicians and students. DACHS: We thank all participants and cooperating clinicians, and Ute Handte‐Daub, Utz Benscheid, Muhabbet Celik and Ursula Eilber for excellent technical assistance. Harvard cohorts (HPFS, NHS, PHS): We would like to thank the participants and staff of the HPFS, NHS and PHS for their valuable contributions as well as the after state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, WY. The authors assume full responsibility for analyses and interpretation of these data. PLCO: The authors thank the PLCO Cancer Screening Trial screening center investigators and the staff from Information Management Services Inc and Westat Inc. Most importantly, we thank the study participants for their contributions that made our study possible. PMH: The authors would like to thank the study participants and staff of the Hormones and Colon Cancer study. WHI: The authors thank the WHI investigators and staff for their dedication, and the study participants for making the program possible. A full listing of WHI investigators can be found at: http://www.whi.org/researchers/Documents%20%20Write%20a%20Paper/WHI%20Investigator%20Short%20List.pdf.
M.T. and E.T. are joint and corresponding authors in this position
Contributor Information
Maria Timofeeva, Email: maria.timofeeva@igmm.ed.ac.uk.
Evropi Theodoratou, Email: e.theodoratou@ed.ac.uk.
References
- 1. Ferlay J, Soerjomataram I, Dikshit R, et al. Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012. Int J Cancer 2015;136:E359–86. [DOI] [PubMed] [Google Scholar]
- 2. Bode AM, Dong Z, Wang H. Cancer prevention and control: alarming challenges in China. Natl Sci Rev 2016;3:117–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Tze CN, Fitzgerald H, Qureshi A, et al. Pioneering annual colorectal cancer screening and treatment targeting Low income communities in Malaysia (20102015). Asian Pac J Cancer Prev 2016;17:3179–83. [PubMed] [Google Scholar]
- 4. Munoz M, Pong‐Wong R, Canela‐Xandri O, et al. Evaluating the contribution of genetics and familial shared environment to common disease using the UKbiobank. Nat Genet 2016;48:980–3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Hutter CM, Mechanic LE, Chatterjee N, et al. Gene‐environment interactions in cancer epidemiology: a National Cancer Institute think tank report. Genet Epidemiol 2013;37:643–57. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Gauderman WJ, Zhang P, Morrison JL, et al. Finding novel genes by testing G x E interactions in a genome‐wide association study. Genet Epidemiol 2013;37:603–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Theodoratou E, Timofeeva M, Li X, et al. Nature, nurture, and cancer risks: genetic and nutritional contributions to cancer. Annu Rev Nutr 2017;37:293–320. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Boffetta P, Winn DM, Ioannidis JP, et al. Recommendations and proposed guidelines for assessing the cumulative evidence on joint effects of genes and environments on cancer occurrence in humans. Int J Epidemiol 2012;41:686–704. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Ioannidis JP, Boffetta P, Little J, et al. Assessment of cumulative evidence on genetic associations: interim guidelines. Int J Epidemiol 2008;37:120–32. [DOI] [PubMed] [Google Scholar]
- 10.World Cancer Research Fund/American Institute for Cancer Research. Diet, Nutrition, Physical Activity and Cancer: a Global Perspective 2018. https://www.wcrf.org/dietandcancer.
- 11.World Cancer Research Fund/American Institute for Cancer Research. Continuous Update Project Report: Diet, Nutrition, Physical Activity and Colorectal Cancer, 2017. https://www.wcrf.org/sites/default/files/Colorectal-Cancer-2017-Report.pdf.
- 12. World Cancer Research Fund . International Systematic Literature Review. The Associations between Food, Nutrition and Physical Activity and the Risk of Colorectal Cancer, 2016. https://www.wcrf.org/sites/default/files/CUP_colorectal_cancer_SLR_2016lo.pdf.
- 13. Bellou V, Belbasis L, Tzoulaki I, et al. Environmental risk factors and Parkinson's disease: an umbrella review of meta‐analyses. Parkinsonism Relat Disord 2016;23:1–9. [DOI] [PubMed] [Google Scholar]
- 14. Welter D, MacArthur J, Morales J, et al. The NHGRI GWAS catalog, a curated resource of SNP‐trait associations. Nucleic Acids Res 2014;42:D1001–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Beck T, Hastings RK, Gollapudi S, et al. GWAS central: a comprehensive resource for the comparison and interrogation of genome‐wide association studies. Eur J Hum Genet 2014;22:949–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Khoury MJ, Bertram L, Boffetta P, et al. Genome‐wide association studies, field synopses, and the development of the knowledge base on genetic variation and human diseases. Am J Epidemiol 2009;170:269–79. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Vineis P, Manuguerra M, Kavvoura FK, et al. A field synopsis on low‐penetrance variants in DNA repair genes and cancer susceptibility. J Natl Cancer Inst 2009;101:24–36. [DOI] [PubMed] [Google Scholar]
- 18. Dimitrakopoulou VI, Travis RC, Shui IM, et al. Interactions between genome‐wide significant genetic variants and circulating concentrations of 25‐Hydroxyvitamin D in relation to prostate cancer risk in the National Cancer Institute BPC3. Am J Epidemiol 2017;185:452–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Theodoratou E, Farrington SM, Tenesa A, et al. Dietary vitamin B6 intake and the risk of colorectal cancer. Cancer Epidemiol Biomarkers Prev 2008;17:171–82. [DOI] [PubMed] [Google Scholar]
- 20. Peters U, Jiao S, Schumacher FR, et al. Identification of genetic susceptibility loci for colorectal tumors in a genome‐wide meta‐analysis. Gastroenterology 2013;144:799–807. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Andersen V, Holst R, Vogel U. Systematic review: diet‐gene interactions and the risk of colorectal cancer. Aliment Pharmacol Ther 2013;37:383–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Andersen V, Vogel U. Systematic review: interactions between aspirin, and other nonsteroidal anti‐inflammatory drugs, and polymorphisms in relation to colorectal cancer. Aliment Pharmacol Ther 2014;40:147–59. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Andersen V, Vogel U. Interactions between meat intake and genetic variation in relation to colorectal cancer. Genes Nutr 2015;10:448. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Corella D, Ordovas JM. Interactions between dietary n‐3 fatty acids and genetic variants and risk of disease. Br J Nutr 2012;107:S271–S83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Cornelis MC. Gene‐coffee interactions and health. Curr Nutr Rep 2014;3:178–95. [Google Scholar]
- 26. Cross JT, Poole EM, Ulrich CM. A review of gene‐drug interactions for nonsteroidal anti‐inflammatory drug use in preventing colorectal neoplasia. Pharmacogenomics J 2008;8:237–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Eichholzer M, Luthy J, Moser U, et al. Folate and the risk of colorectal, breast and cervix cancer: the epidemiological evidence. Swiss Med Wkly 2001;131:539–49. [DOI] [PubMed] [Google Scholar]
- 28. Houlston RS, Tomlinson IP. Polymorphisms and colorectal tumor risk. Gastroenterology 2001;121:282–301. [DOI] [PubMed] [Google Scholar]
- 29. Karahalil B, Bohr VA, Wilson DM 3rd. Impact of DNA polymorphisms in key DNA base excision repair proteins on cancer risk. Hum Exp Toxicol 2012;31:981–1005. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Klarich DS, Brasser SM, Hong MY. Moderate alcohol consumption and colorectal cancer risk. Alcohol Clin Exp Res 2015;39:1280–91. [DOI] [PubMed] [Google Scholar]
- 31. Kostner K, Denzer N, Muller CS, et al. The relevance of vitamin D receptor (VDR) gene polymorphisms for cancer: a review of the literature. Anticancer Res 2009;29:3511–36. [PubMed] [Google Scholar]
- 32. Lenihan‐Geels G, Bishop KS, Ferguson LR. Cancer risk and eicosanoid production: interaction between the protective effect of long chain Omega‐3 polyunsaturated fatty acid intake and genotype. J Clin Med 2016;5(2):pii:E25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Sharp L, Little J. Polymorphisms in genes involved in folate metabolism and colorectal neoplasia: a HuGE review. Am J Epidemiol 2004;159:423–43. [DOI] [PubMed] [Google Scholar]
- 34. Shin A, Kim J. Effect modification of meat intake by genetic polymorphisms on colorectal neoplasia susceptibility. Asian Pac J Cancer Prev 2010;11:281–7. [PubMed] [Google Scholar]
- 35. Simonds NI, Ghazarian AA, Pimentel CB, et al. Review of the gene‐environment interaction literature in cancer: what do we know? Genet Epidemiol 2016;40:356–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Ananthakrishnan AN, Du M, Berndt SI, et al. Red meat intake, NAT2, and risk of colorectal cancer: a pooled analysis of 11 studies. Cancer Epidemiol Biomarkers Prev 2015;24:198–205. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Ding W, Zhou DL, Jiang X, et al. Methionine synthase A2756G polymorphism and risk of colorectal adenoma and cancer: evidence based on 27 studies. PLoS One 2013;8:e60508. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Hiraki LT, Joshi AD, Ng K, et al. Joint effects of colorectal cancer susceptibility loci, circulating 25‐hydroxyvitamin D and risk of colorectal cancer. PLoS One 2014;9:e92212. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Hutter CM, Chang‐Claude J, Slattery ML, et al. Characterization of gene‐environment interactions for colorectal cancer susceptibility loci. Cancer Res 2012;72:2036–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Hutter CM, Slattery ML, Duggan DJ, et al. Characterization of the association between 8q24 and colon cancer: gene‐environment exploration and meta‐analysis. BMC Cancer 2010;10:670. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Jiao S, Hsu L, Bezieau S, et al. SBERIA: set‐based gene‐environment interaction test for rare and common variants in complex diseases. Genet Epidemiol 2013;37:452–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Kantor ED, Hutter CM, Minnier J, et al. Gene‐environment interaction involving recently identified colorectal cancer susceptibility loci. Cancer Epidemiol Biomarkers Prev 2014;23:1824–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Liu Y, Qin H, Zhang Y, et al. P53 codon 72 polymorphism and colorectal cancer: a meta‐analysis of epidemiological studies. Hepatogastroenterology 2011;58:1926–9. [DOI] [PubMed] [Google Scholar]
- 44. Nan H, Morikawa T, Suuriniemi M, et al. Aspirin use, 8q24 single nucleotide polymorphism rs6983267, and colorectal cancer according to CTNNB1 alterations. J Natl Cancer Inst 2013;105:1852–61. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Pabalan N, Bapat B, Sung L, et al. Cyclin D1 Pro241Pro (CCND1‐G870A) polymorphism is associated with increased cancer risk in human populations: a meta‐analysis. Cancer Epidemiol Biomarkers Prev 2008;17:2773–81. [DOI] [PubMed] [Google Scholar]
- 46. Pabalan N, Jarjanazi H, Ozcelik H. A meta‐analysis of the C1420T polymorphism in cytosolic serine hydroxymethyltransferase (SHMT1) among Caucasian colorectal cancer populations. Int J Colorectal Dis 2013;28:925–32. [DOI] [PubMed] [Google Scholar]
- 47. Raimondi S, Botteri E, Iodice S, et al. Gene‐smoking interaction on colorectal adenoma and cancer risk: review and meta‐analysis. Mutat Res 2009;670:6–14. [DOI] [PubMed] [Google Scholar]
- 48. Smits KM, Gaspari L, Weijenberg MP, et al. Interaction between smoking, GSTM1 deletion and colorectal cancer: results from the GSEC study. Biomarkers 2003;8:299–310. [DOI] [PubMed] [Google Scholar]
- 49. Tse G, Eslick GD. Cruciferous vegetables and risk of colorectal neoplasms: a systematic review and meta‐analysis. Nutr Cancer 2014;66:128–39. [DOI] [PubMed] [Google Scholar]
- 50. Wan H, Zhou Y, Yang P, et al. Genetic polymorphism of glutathione S‐transferase T1 and the risk of colorectal cancer: a meta‐analysis. Cancer Epidemiol 2010;34:66–72. [DOI] [PubMed] [Google Scholar]
- 51. Wang H, Iwasaki M, Haiman CA, et al. Interaction between red meat intake and NAT2 genotype in increasing the risk of colorectal cancer in Japanese and African Americans. PLoS One 2015;10:e0144955. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 52. Wang X, Xing GH, Fan CC. Association between the FAS rs2234767G/a polymorphism and cancer risk: a systematic review and meta‐analysis. DNA Cell Biol 2014;33:320–7. [DOI] [PubMed] [Google Scholar]
- 53. Lq Z, Jn Z, Wang J, et al. Absence of association between N‐Acetyltransferase 2 acetylator status and colorectal cancer susceptibility: based on evidence from 40 studies. PLoS One 2012;7(3):e32425. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54. Danfeng S, Xia W, Jingyuan F. Meta‐analysis of the relationship between methylenetetrahydrofolate reductase gene polymorphism and colon cancer. Gastroenterology 2006;11:516–21. [Google Scholar]
- 55. Figueiredo JC, Hsu L, Hutter CM, et al. Genome‐wide diet‐gene interaction analyses for risk of colorectal cancer. PLoS Genet 2014;10:e1004228. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 56. Figueiredo JC, Lewinger JP, Song C, et al. Genotype‐environment interactions in microsatellite stable/microsatellite instability‐low colorectal cancer: results from a genome‐wide association study. Cancer Epidemiol Biomarkers Prev 2011;20:758–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 57. Nan H, Hutter CM, Lin Y, et al. Association of aspirin and NSAID use with risk of colorectal cancer according to genetic variants. JAMA 2015;313:1133–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 58. Du M, Zhang X, Hoffmeister M, et al. No evidence of gene‐calcium interactions from genome‐wide analysis of colorectal cancer risk. Cancer Epidemiol Biomarkers Prev 2014;23:2971–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 59. Garcia‐Albeniz X, Rudolph A, Hutter C, et al. CYP24A1 variant modifies the association between use of oestrogen plus progestogen therapy and colorectal cancer risk. Br J Cancer 2016;114:221–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 60. Gong J, Hutter CM, Newcomb PA, et al. Genome‐wide interaction analyses between genetic variants and alcohol consumption and smoking for risk of colorectal cancer. PLoS Genet 2016;12:e1006296. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 61. Jiao S, Peters U, Berndt S, et al. Powerful set‐based gene‐environment interaction testing framework for complex diseases. Genet Epidemiol 2015;39:609–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 62. Siegert S, Hampe J, Schafmayer C, et al. Genome‐wide investigation of gene‐environment interactions in colorectal cancer. Hum Genet 2013;132:219–31. [DOI] [PubMed] [Google Scholar]
- 63. Qiao Y, Yang T, Gan Y, et al. Associations between aspirin use and the risk of cancers: a meta‐analysis of observational studies. BMC Cancer 2018;18:288. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 64. Tanskanen T, van den Berg L, Valimaki N, et al. Genome‐wide association study and meta‐analysis in northern European populations replicate multiple colorectal cancer risk loci. Int J Cancer 2017:142(3):540–546. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 65. Orlando G, Law PJ, Palin K, et al. Variation at 2q35 (PNKD and TMBIM1) influences colorectal cancer risk and identifies a pleiotropic effect with inflammatory bowel disease. Hum Mol Genet 2016;25:2349–59. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 66. Lin KJ, Cheung WY, Lai JY, et al. The effect of estrogen vs combined estrogen‐progestogen therapy on the risk of colorectal cancer. Int J Cancer 2012;130:419–30. [DOI] [PubMed] [Google Scholar]
- 67. Schumacher FR, Schmit SL, Jiao S, et al. Genome‐wide association study of colorectal cancer identifies six new susceptibility loci. Nat Commun 2015;6:7138. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 68. Pomerantz MM, Ahmadiyeh N, Jia L, et al. The 8q24 cancer risk variant rs6983267 shows long‐range interaction with MYC in colorectal cancer. Nat Genet 2009;41:882–4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 69. Tuupanen S, Turunen M, Lehtonen R, et al. The common colorectal cancer predisposition SNP rs6983267 at chromosome 8q24 confers potential to enhanced Wnt signaling. Nat Genet 2009;41:885–90. [DOI] [PubMed] [Google Scholar]
- 70. Sur IK, Hallikas O, Vaharautio A, et al. Mice lacking a Myc enhancer that includes human SNP rs6983267 are resistant to intestinal tumors. Science (New York, NY) 2012;338:1360–3. [DOI] [PubMed] [Google Scholar]
- 71. Bos CL, Kodach LL, van den Brink GR, et al. Effect of aspirin on the Wnt/beta‐catenin pathway is mediated via protein phosphatase 2A. Oncogene 2006;25:6447–56. [DOI] [PubMed] [Google Scholar]
- 72. Fresno Vara JA, Casado E, de Castro J, et al. PI3K/Akt signalling pathway and cancer. Cancer Treat Rev 2004;30:193–204. [DOI] [PubMed] [Google Scholar]
- 73. Kaur J, Sanyal SN. PI3‐kinase/Wnt association mediates COX‐2/PGE(2) pathway to inhibit apoptosis in early stages of colon carcinogenesis: chemoprevention by diclofenac. Tumour Biol 2010;31:623–31. [DOI] [PubMed] [Google Scholar]
- 74. Grivennikov SI, Karin M. Inflammatory cytokines in cancer: tumour necrosis factor and interleukin 6 take the stage. Ann Rheum Dis 2011;70(Suppl 1):i104–8. [DOI] [PubMed] [Google Scholar]
- 75. Klampfer L. Cytokines, inflammation and colon cancer. Curr Cancer Drug Targets 2011;11:451–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 76. Symmons O, Uslu VV, Tsujimura T, et al. Functional and topological characteristics of mammalian regulatory domains. Genome Res 2014;24:390–400. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 77. Bareis P, Bises G, Bischof MG, et al. 25‐hydroxy‐vitamin d metabolism in human colon cancer cells during tumor progression. Biochem Biophys Res Commun 2001;285:1012–7. [DOI] [PubMed] [Google Scholar]
- 78. Dong LM, Ulrich CM, Hsu L, et al. Vitamin D related genes, CYP24A1 and CYP27B1, and colon cancer risk. Cancer Epidemiol Biomarkers Prev 2009;18:2540–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 79. Hosoya T, Maillard I, Engel JD. From the cradle to the grave: activities of GATA‐3 throughout T‐cell development and differentiation. Immunol Rev 2010;238:110–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 80. Hedlund M, Padler‐Karavani V, Varki NM, et al. Evidence for a human‐specific mechanism for diet and antibody‐mediated inflammation in carcinoma progression. Proc Natl Acad Sci U S A 2008;105:18936–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 81. Zheng R, Blobel GA. GATA transcription factors and cancer. Genes Cancer 2010;1:1178–88. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 82. Schlessinger A, Matsson P, Shima JE, et al. Comparison of human solute carriers. Protein Sci 2010;19:412–28. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 83. Hoglund PJ, Nordstrom KJ, Schioth HB, et al. The solute carrier families have a remarkably long evolutionary history with the majority of the human families present before divergence of Bilaterian species. Mol Biol Evol 2011;28:1531–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 84. Smith PG, Day NE. The design of case‐control studies: the influence of confounding and interaction effects. Int J Epidemiol 1984;13:356–65. [DOI] [PubMed] [Google Scholar]
- 85. Thomas D. Gene‐‐environment‐wide association studies: emerging approaches. Nat Rev Genet 2010;11:259–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 86. Burton PR, Hansell AL, Fortier I, et al. Size matters: just how BIG is BIG?: quantifying realistic sample size requirements for human genome epidemiology. Int J Epidemiol 2009;38:263–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 87. Cornelis MC, Hu FB. Gene‐environment interactions in the development of type 2 diabetes: recent progress and continuing challenges. Annu Rev Nutr 2012;32:245–59. [DOI] [PubMed] [Google Scholar]
- 88. Theodoratou E, Tzoulaki I, Zgaga L, et al. Vitamin D and multiple health outcomes: umbrella review of systematic reviews and meta‐analyses of observational studies and randomised trials. BMJ (Clin Res Ed) 2014;348:g2035. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 89. Belbasis L, Savvidou MD, Kanu C, et al. Birth weight in relation to health and disease in later life: an umbrella review of systematic reviews and meta‐analyses. BMC Med 2016;14:147. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 90. Sohani ZN, Meyre D, de Souza RJ, et al. Assessing the quality of published genetic association studies in meta‐analyses: the quality of genetic studies (Q‐genie) tool. BMC Genet 2015;16:50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 91. Li X, Meng X, Timofeeva M, et al. Serum uric acid levels and multiple health outcomes: umbrella review of evidence from observational studies, randomised controlled trials, and Mendelian randomisation studies. BMJ (Clin Res Ed) 2017;357:j2376. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 92. Borenstein M, Higgins JP, Hedges LV, et al. Basics of meta‐analysis: I(2) is not an absolute measure of heterogeneity. Res Synth Methods 2017;8:5–18. [DOI] [PubMed] [Google Scholar]
- 93. Geneletti S, Gallo V, Porta M, et al. Assessing causal relationships in genomics: from Bradford‐Hill criteria to complex gene‐environment interactions and directed acyclic graphs. Emerg Themes Epidemiol 2011;8:5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 94. Fedak KM, Bernal A, Capshaw ZA, et al. Applying the Bradford Hill criteria in the 21st century: how data integration has changed causal inference in molecular epidemiology. Emerg Themes Epidemiol 2015;12:14. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.